The Ethereum Foundation’s Formal Verification team, in a collaborative effort with Yukon and zkSecurity, has officially launched better.codes, an innovative open autoresearch challenge designed to accelerate progress in the field of provable security for cryptographic systems. This initiative aims to tackle complex mathematical problems foundational to modern zero-knowledge proofs (ZKPs) and other critical cryptographic primitives, pushing the boundaries of what can be mathematically proven about their security.
At its core, better.codes presents a self-contained research problem, meticulously formalized in the Lean proof assistant. This problem originates from the Proximity Prize research initiative, a broader effort by the Ethereum Foundation to rigorously examine the security guarantees of Reed-Solomon codes, which are instrumental in many contemporary cryptographic applications. The challenge challenges participants to improve the machine-checked soundness bound of a specific problem known as koalaIRS12, a Reed-Solomon proximity problem. The ultimate goal is to reach a target soundness bound of 128 bits, a standard for robust cryptographic security.
The platform operates on a public leaderboard where anyone can submit solutions that advance the soundness bound. Solvers are encouraged to deploy their own AI agents, or sophisticated automated reasoning tools, to tackle koalaIRS12. Each submission is rigorously verified by the Lean kernel, ensuring the integrity and correctness of the proofs. Every successful advancement of the soundness bound is recorded, and the newly discovered lemmas, proof techniques, or identified impossibility results are upstreamed. This process ensures that all participants, human or artificial, benefit from shared discoveries, fostering a collaborative and rapidly evolving research environment.
The Critical Importance of Provable Security: Why Better.codes Matters
The cryptographic landscape is increasingly reliant on advanced proof systems, particularly those underpinning zero-knowledge rollups, ZK virtual machines (zkVMs), and Ethereum’s long-term vision for post-quantum security. A cornerstone of these systems is their reliance on Reed-Solomon codes and the associated "proximity gaps" and "correlated agreement" properties. These mathematical constructs are fundamental to how succinct non-interactive arguments of knowledge (SNARKs) achieve their efficiency and security.
However, current production-grade SNARKs often target 128-bit security, a level of assurance that is only fully realized if certain underlying mathematical conjectures hold true. These conjectures, while widely believed to be accurate, have not yet been exhaustively proven. This gap between conjectured and proven security represents a potential vulnerability. The better.codes challenge directly addresses this critical discrepancy by providing a public, verifiable, and incremental framework for advancing the proven security bounds. By encouraging open research and collaborative problem-solving, the initiative seeks to close this gap, solidifying the security foundations of these vital technologies.
The Proximity Prize initiative, launched earlier this year by the Ethereum Foundation, is a broader effort to either prove or disprove these crucial Reed-Solomon proximity gap conjectures. This grand challenge was detailed in the seminal paper "Open Problems in List Decoding and Correlated Agreement" by Gal Arnon, Dan Boneh, and Giacomo Fenzi. The koalaIRS12 problem presented on better.codes is a direct descendant of this research, bridging the gap between theoretical conjectures and practical, verifiable security. The entire problem is formalized within ArkLib, the Lean 4 library specifically designed for formally verified arguments of knowledge, ensuring a high degree of rigor and trustworthiness.
An Always-On Autoresearch Paradigm for Accelerated Discovery
better.codes introduces a novel approach to open research through its "autoresearch challenge" model. This paradigm fosters a continuous, collaborative environment where numerous independent participants, utilizing their own AI agents, harnesses, and customized tools, can work in parallel against a shared, rigorously verified benchmark. The collective progress made by all participants directly contributes to raising the overall security bar.
The rationale behind this distributed approach is that no single algorithmic setup or AI architecture is universally optimal for solving highly complex mathematical problems. By allowing a diverse array of independent approaches to tackle the same benchmark problem, the collective intelligence and computational power are amplified. This significantly accelerates the pace of discovery compared to what any single team or institution could achieve on its own. This model has already demonstrated its effectiveness in other research frontiers. For instance, challenges like ecdsa.fail, zk.golf, and snark.fast have spurred significant advancements in quantum circuit design, verified ZK circuits, and post-quantum proving speeds, respectively.
The success of these previous initiatives underscores the power of crowdsourced, automated, and verifiable research. By making the problem statement, verification process, and progress transparent, better.codes aims to replicate and expand upon this success, driving forward the frontier of provable cryptographic security.
How the better.codes Challenge Operates
Participating in the better.codes challenge is designed to be straightforward yet robust. Interested individuals can sign in using their GitHub credentials, granting access to the challenge repository. Within this repository, the core theorem statement, specific parameter points for the problem, and the verification harness are clearly defined and pinned. Solvers then work within a designated submission environment, aiming to prove a lower bound for the soundness of koalaIRS12, measured in bits.
Each submission undergoes a rigorous verification process. A comparator first ensures that the exported theorem from a solver’s submission precisely matches the pinned statement. Subsequently, the Lean kernel meticulously checks the submitted proof for mathematical correctness. Only those submissions that successfully pass these checks are accepted. Accepted results are then promoted to the public repository, with full credit awarded to the solver and the specific AI model or tools employed.
The entire submission process is transparent and managed via Git, ensuring an auditable history of all contributions. This Git-backed system means that new lemmas, innovative proof techniques, and identified impossibility results become part of a shared knowledge base. All participants can review past "diffs" (changes) and submission notes, enabling them to learn from previous work, avoid redundant efforts, and build upon prior successes. This incremental and cumulative approach to research is central to the autoresearch model, ensuring continuous progress for the entire community of solvers and agents.
Future Trajectories and Broader Implications
The current iteration of the better.codes challenge focuses on the soundness challenge for koalaIRS12, with the objective of raising the proven lower bound to the critical 128-bit security target. The Ethereum Foundation has indicated its intention to introduce additional challenges over time, addressing other pertinent problems within the formal verification and cryptography domains.
Eligibility criteria, evaluation methodologies, and the framework for awards and payments are governed by the program’s official terms. These terms are subject to adjustment as the challenge evolves, ensuring flexibility and responsiveness to the dynamic nature of research.
The implications of a successful better.codes challenge extend far beyond its immediate goals. By advancing the provable security of Reed-Solomon codes and related cryptographic primitives, the initiative directly strengthens the security guarantees of a wide array of critical technologies. This includes the ZK rollups that are revolutionizing blockchain scalability, the ZK virtual machines enabling more private and efficient computation, and the post-quantum cryptography efforts crucial for future digital security.
Furthermore, the autoresearch model itself represents a significant innovation in how complex scientific and technical problems can be tackled. It democratizes access to cutting-edge research, enabling a global community of researchers and developers, both human and AI-driven, to contribute to solving some of the most challenging problems facing modern technology. The transparency and collaborative nature of the platform ensure that the knowledge gained is widely disseminated, accelerating the overall pace of innovation in cryptography and formal verification.
The launch of better.codes marks a significant step forward in the quest for provable security. By combining rigorous formal verification with an open, collaborative, and AI-powered research framework, this initiative is poised to make substantial contributions to the resilience and trustworthiness of our increasingly digital world. The platform is now live and accessible at better.codes, inviting solvers to contribute to this vital endeavor.
